Moonfire

Our October newsletter is scarily good. It's been a stellar few weeks for our portfolio founders: Filigran closed its $35M Series B; Lightdash announced a $11M Series A; and for LiveFlow – a $13.5M Series A. Congratulations all. And, this month, our feature looks at the rise of the 'venture creator' in conversation with Harry Stebbings, creator of 20VC, and Jordan Schwarzenberger, Manager of The Sidemen. We also have our usual podcast and book recommendations, and snapshot of interesting news and ideas we've come across. Read & subscribe: https://lnkd.in/eDCwJjJZ Happy Halloween from Moonfire 🎃

Our first investment in Filigran was in early 2023. Then, back in February, we participated in its $15M Series A. And only 6 months later, they've raised a $35M Series B. Samuel Hassine and Julien Richard's founding story is an incredible one. They've scaled from two to 80 people in 2 years, won blue chip clients such as Airbus, Marriott and Hermès, and opened an office in the US. It's been a privilege to support them from the very beginning. Congratulations Samuel and Julien 👏 We interviewed Samuel back in February when they announced the Series A, and you can read about his and Julien's vision and journey here: https://lnkd.in/d5TRMha9

The world’s fastest text-to-speech solution has an ultra low latency of just 25 milliseconds. Meet Neuphonic – a Voice AI with a patent-pending algorithm that enables real time incremental speech generation. This is a game-changing technology that could enable faster, more natural interactions in customer service, create genuinely immersive gaming experiences, and significantly increase accessibility across so many industries as companies use Neuphonic’s API to create human-like speech in their products. Neuphonic was created by two of the smartest founders we’ve ever met: Sohaib A., a former hedge fund quant trader, and Jiameng Gao, co-founder of speech-to-speech translation Papercup. They are both multilingual Cambridge ML MPhil graduates and 1st gen immigrants attuned to language barriers, the perfect people to lead on solving text-to-speech challenges. Voice AI is a market that's set to grow USD 41.39 billion by 2030, and Neuphonic are innovating right at the forefront of it. We’re incredibly excited to have led their £3m pre-seed alongside Tiny Supercomputer Investment Company, Salica Investments and Cur8 Capital, to support Sohaib and Jiameng as they refine and release their tech, grow their team and work towards global scalability.

https://lnkd.in/eWZ-UCxn

Just over five years ago, I met Lightdash’s now co-founder and CTO Oliver Laslett to try and offer him a job at Moonfire. But when he told me he was starting a company and had been accepted to YC I offered to invest instead. Oliver and co-founder and CEO Hamzah Chaudhary graduated from YC's S20 cohort with a tool called Hubble, which helped companies identify issues with data quality. But when they realised this would be most useful inside a BI tool, they pivoted, focusing instead on empowering data analytics teams with the tools to be as productive as software engineers. Lightdash was born: an open-source business intelligence platform for modern data teams. Fast forward to today, Lightdash has announced their $11M Series A led by Accel with Shopify Ventures, Operator Partners and Y Combinator, plus Moonfire. Lightdash is now building a whole new way to self-serve data with the launch of an AI Analyst service, which allows users to chat with their data in natural language and receive relevant curated insights. Congratulations Oliver and Hamzah – there are more great things to come. https://lnkd.in/eNKavXAz

Education and learning have never been more divergent than they are today. It feels like despite billions of dollars invested into the space, education delivery is still solving for the lowest common denominator without care for evolving student expectations, personalised learning pathways and modern information delivery. In Revyze, we see a solution that taps into the needs of today's student demography - social at heart, personalised by design, truly gamified, accessible at your fingertips and authentically taught by other students, relevant educators and AI. Most importantly, we are super excited to support Florent Sciberras, Guillaume Perrot and Anatole Blanc as they take great strides towards closing the chasm between education and learning in a way that is fun for students. We at Moonfire are happy to have co-led their seed round along with our friends at Speedinvest. 🚀 Mattias Ljungman Jeroen Arts Maximilian Wilhelm Sameer Singh You can find out more about Revyze via Pulse, in the comments below. ⬇️

🇺🇸 🇺🇸 A year ago, Revyze reached #1 of the app store in France. Guillaume Perrot took a flight to NYC to expand Revyze in the US. 3 months later, in January, we were failing, Revyze was in crisis. Our product was not ready for global expansion. So we paused the US, went back to building our core product, iterated, relaunched, gained 20 retention points, and grew monthly active users 20x in 3 months. We reached over 1 million users & one-third of 9th graders in France learned on Revyze in June. 🚀🚀 Today, we're incredibly excited to share that Revyze raised a 5.5M€ (6M$) seed round, led by Speedinvest and Moonfire to build the first global social learning app for GenZ and GenAlpha. The round includes participation Ilkka Paananen (Supercell cofounder), Sebastian Knutsson and Riccardo Zacconi (King / Candycrush cofounders), Nickey Skarstad (Director of Product, Duolingo), Motier Ventures and STATION F - all of whom we're so excited to learn from. Why we wake up every day is for all the students that tell us they are bored and stressed in school. They tell us "I learn better in 2 minutes on Revyze than in 4 hours in class" on Revyze because "it's easy to use & understand", "they can learn when they want and where they want" and "it's less stressful than school". This new funding will help us accelerate on a growing trend we've observed by calling our users over the summer. They told us: "every morning, I open Insta and Revyze". We want to develop a daily habit so that you come on Revyze every day to play, learn, even if there is no test around the corner. We will do this by making the learning experience more gamified (think Duolingo), and more social (think Tiktok). Revyze will have achieved its mission when where you come from no longer determines where you can go. The road to achieve our mission is incredibly steep and full of pitfalls. We will take risks. And fail a thousand times more. And learn a thousand times. We’re just getting started. And we couldn’t be more excited for the road ahead to help everyone learn better, together 💙💙 (TechCrunch article in comment)

What if we could turn teenagers’ social media scrolling habits into something more productive? Revyze is on a mission to transform education and inspire Gen Z and Gen Alpha towards a passion for lifelong learning. 🎓 They’ve built the world’s first global social learning app – short-form videos and quiz content, all of which is moderated and verified. The three co-founders Guillaume Perrot (ex-Alan), Anatole Blanc (ex-Auto1), and Florent Sciberras (ex-Airbnb) started working on Revyze while they were in college. In two months, they had built and launched an app that hit 40,000 downloads in just a few days. Now, they have over 1 million users. It was a very exciting opportunity to co-lead Revyze’s $6m seed round with Speedinvest, alongside Ilkka Paananen (cofounder of Supercell), Riccardo Zacconi and Sebastian Knutsson (cofounders of King / Candycrush), Nickey Skarstad (Director of Product, Duolingo) and STATION F, among others. You can find out more about Revyze via Pulse, in the comments below. ⬇️

In the age of ever more capable and knowledgeable LLMs, how best to leverage proprietary data? RAG (retrieval-augmented generation) has become the go-to strategy and many AI products today use a RAG pipeline under the hood. There are many levers you can pull to augment a vanilla RAG pipeline - e.g. tweaking hyperparameters like size of chunks or chunk overlap, trying out different post-retrieval techniques like re-ranking or even agentic systems. But there is also fine-tuning. With OpenAI's recent fine-tuning capabilities it's easier than ever to try it out on your proprietary data and quickly see if it leads to an improvement. If it is, and you don't mind the extra effort/cost, you can always try out open-source LLMs afterwards. In this blog post we explore the following: can we improve the performance of our RAG pipeline by fine-tuning an LLM and leaving all other components untouched? Spoiler alert: we can. For more details read our latest blog post here: https://lnkd.in/eRedCcVX

What is the future of play in the age of infinite content? 🎮 How can AI augment human creativity, not replace it? Who is the author? And how can AI redefine what we even consider a game to be? Thank you to Suveer Kothari, Gunnar Holmsteinn, Nihal Tharoor, and Benedict Tatham for reading drafts of our latest piece, now live in our September newsletter: The Future of Play. Read and subscribe: https://lnkd.in/euNZBU3Y #gaming #AI #LLMs